EFSA CONTAM 2009 — Arsenic in Food

The European Food Safety Authority Panel on Contaminants in the Food Chain Scientific Opinion on arsenic in food, adopted 12 October 2009 in response to Question No EFSA-Q-2008-425 from the European Commission. It assessed risks from inorganic arsenic in food using more than 100,000 occurrence data points submitted by 15 European Member States plus Norway covering 2003 to 2008, and concluded that the long-standing Joint FAO/WHO Expert Committee on Food Additives provisional tolerable weekly intake of 15 µg inorganic arsenic per kilogram body weight per week is no longer appropriate because adverse effects, including cancers of the lung and urinary bladder, have been documented at exposures below the PTWI. In place of a single tolerable intake, the Panel identified a range of BMDL01 values of 0.3 to 8 µg inorganic arsenic per kilogram body weight per day for skin lesions and cancers of the lung, urinary bladder and skin, and recommended that risk be characterised using margins of exposure between this range and estimated dietary exposures. Across the 19 European countries with food consumption data in the EFSA Concise European Food Consumption Database, mean inorganic arsenic exposure for adults was estimated at 0.13 to 0.56 µg/kg b.w./day and 95th percentile exposure at 0.37 to 1.22 µg/kg b.w./day (scenario 5: 70 percent inorganic arsenic in non-fish/non-seafood food categories, fixed values of 0.03 mg/kg in fish and 0.1 mg/kg in seafood, lower-bound to upper-bound). Because these adult exposures fall within the BMDL01 range, the Panel concluded that there is little or no margin of exposure and the possibility of risk to some consumers cannot be excluded. Children under three years of age were estimated to be exposed at 2 to 3 times the per-kg adult rate (0.50 to 2.66 µg/kg b.w./day), and high consumers of rice or algae-based products can reach 1 to 4 µg/kg b.w./day from those food sources alone. The opinion supersedes the JECFA 1989 PTWI for the EU and is the EU reference assessment for inorganic arsenic in food until the 2021 chronic-exposure update (EFSA Journal 2021;19(1):6380, see efsa2021-chronic-exposure-iAs).

This page covers the published EFSA Journal 1351 record (199 pp.). The published file is the corrected version dated 1 February 2010 that fixed an error in Rahman et al. (2006a) skin-lesion data in Table 41; the underlying conclusions are unchanged from the 22 October 2009 original publication. The opinion was adopted by the Panel on 12 October 2009 and first published 22 October 2009; the corrected file replaces that publication, not the adoption.

Key numbers

Reference points for inorganic arsenic (Section 9, “Conclusions”):

Reference points by endpoint and population (Table 43, p. 141):

BMDL01 ranges are the Panel-derived dietary equivalents assuming 3 to 5 L/day water plus 50 to 200 µg/day iAs from food at 55 kg body weight for rural Asian populations, or 1 to 2 L/day water plus 10 to 20 µg/day iAs from food at 70 kg body weight for North/South American populations (Section 8.4.1.1). The 0.3 µg/kg b.w./day lower bound of the overall range is anchored on the Mongolian and New Hampshire data; the 8 µg/kg b.w./day upper bound comes from the Bangladesh dermal-lesion and Taiwanese bladder-cancer studies.

Dietary exposure to total and inorganic arsenic across the 19 European countries with consumption data in the Concise European Food Consumption Database (Tables 23, 25, 26; scenario 5 used as central iAs scenario):

LB and UB are lower and upper bounds for treatment of non-detects (LB sets <LOD to 0, UB sets <LOD to LOD or LOQ). Scenario 5 fixes inorganic-As at 0.03 mg/kg in fish and 0.1 mg/kg in seafood, and assumes 70 percent of total As is inorganic in the seven main contributing non-fish food categories (cereals, vegetables/nuts/pulses, fruit and vegetable juices and soft drinks and bottled water, coffee/tea/cocoa, alcoholic beverages, miscellaneous and special dietary uses, tap water). Sensitivity across the nine inorganic-As scenarios (Table 25) is modest because the fish/seafood iAs assumption only marginally affects iAs totals; the dominant driver is the 50/70/100 percent assumption for the non-fish categories.

Dietary inorganic arsenic exposure in European children by age, scenario 5, beta-binomial-normal model across 3 to 10 Member States (Table 28, p. 69):

Infant exposure to inorganic arsenic at 6 months and 6.1 kg body weight, scenario 5 (Table 27, p. 67):

Specialised-diet exposure increments added on top of the adult scenario 5 base of 0.43 µg iAs/kg b.w./day (Table 31, p. 74):

Adjusted mean total-As occurrence in foods used as exposure inputs (Table 21, p. 53, mg/kg LB–UB):

Inorganic-arsenic proportions of total As assumed or measured (Section 5.5, p. 56–57):

Major dietary contributors to inorganic-arsenic exposure in the general European population (Figure 10, Section 7.3.1, p. 66):

Occurrence dataset composition (Section 5.2, p. 33–37):

Methods (brief)

The opinion is a risk assessment combining occurrence data, consumption data, hazard identification and dose-response modelling.

Occurrence data were solicited under DATEX-2008-0012 and totalled 100,867 records from 14 EU Member States plus Norway; 77,275 records survived the data-cleaning step (which excluded pre-2003 samples, records with insufficient analytical detail, and laboratories reporting unacceptably high LODs >0.1 mg/kg or LOQs >0.3 mg/kg). Analytical techniques in the contributing laboratories included hydride-generation atomic absorption spectrometry (HG-AAS, most common at 39 percent), electrothermal AAS (ET-AAS), inductively coupled plasma mass spectrometry (ICP-MS) and minority use of HG-AFS and HG-ICP-AES; 39 percent of records did not specify the analytical method. Because two thirds of records were below LOD, all category means were computed in both lower-bound and upper-bound non-detect treatments per WHO GEMS/Food convention (LB sets <LOD to zero, UB sets <LOD to LOD/LOQ). Analytical performance characteristics for total arsenic are legislated only for drinking water; for foods there is no harmonised LOD/LOQ requirement, contributing to wide between-laboratory spread.

Consumption data were drawn from the EFSA Concise European Food Consumption Database (19 contributing countries, individual-level records used to derive country-specific means and 95th percentiles). Child consumption data were supplied by a separate EFSA-contracted consortium (University of Ghent coordination) covering 11 European countries with food consumption records for children aged 1 to 14 years; long-term inorganic-arsenic exposure was modelled with the beta-binomial-normal (BBN) model in the MCRA software. Italian children were additionally compared to Italian adults using the 1994–1996 INRAN national survey of 1,940 subjects.

Because only 919 records reported inorganic arsenic directly, the Panel constructed nine inorganic-arsenic exposure scenarios (Table 24, p. 64) by crossing three fish/seafood iAs assumptions (measured LB/UB; fixed 0.03/0.10 mg/kg; fixed 0.015/0.05 mg/kg) with three non-fish/non-seafood iAs assumptions (50, 70 or 100 percent of total). Scenario 5 (70 percent in non-fish; fixed 0.03/0.10 mg/kg in fish/seafood) was carried forward as the central estimate because it sat at the median of the nine scenarios for both mean and P95 adult exposure.

Hazard characterisation drew on epidemiological cohort and case-control studies of dermal lesions, skin cancer, lung cancer and urinary bladder cancer in populations exposed to arsenic-contaminated drinking water (Bangladesh, Mongolia, Chile, North-East Taiwan, New Hampshire). Animal toxicity data were judged insufficiently representative of human dose-response and were not used as the basis for the reference points. Benchmark dose modelling was performed in US EPA BMDS version 2.0 with a benchmark response of 1 percent extra risk; for each study the lower 95 percent confidence limit of the benchmark concentration in drinking water (BMCL01) was computed, then converted to a dietary BMDL01 by adding assumed direct water consumption (1–2 L/day for North America, 3–5 L/day for rural Asia including cooking water) plus iAs from food (10–20 µg/day for North America, 50–200 µg/day for rural Asia per the US EPA Science Advisory Board sensitivity range), divided by assumed body weight (70 kg for North America, 55 kg for rural Asia). Because the conversion from water concentration to total dietary exposure carries substantial assumption-driven uncertainty, the Panel reported a range of BMDL01 values across endpoints and study populations (0.3 to 8 µg/kg b.w./day) rather than a single reference point.

The opinion explicitly excluded organic arsenic from the risk characterisation, with the exceptions: arsenobetaine (the major form in fish and most seafood) is not metabolised in humans and is assumed to be of no toxicological concern; arsenosugars and arsenolipids are metabolised to dimethylarsinate but human toxicity data are absent; methylarsonate and dimethylarsinate were not characterised further owing to lack of European occurrence data. Body weights of 60 kg (adults, general) and 6.1 kg (6-month-old infants, derived from German DONALD study) were used in the EU exposure calculations.

Uncertainty was scored qualitatively (Table 44, p. 145) across ten sources including measurement uncertainty, geographical extrapolation, sampling design, non-detect treatment, iAs/tAs assumptions, food processing, single-cause attribution and lack of animal models. The Panel judged the combined effect of using upper-bound exposure with lower-bound BMDL01 to be conservative.

Implications

• Reference value: this opinion establishes the EU determination that the 1989 JECFA PTWI of 15 µg/kg b.w./week for inorganic arsenic is no longer health-protective and replaces a tolerable-intake framework with a margin-of-exposure framework anchored on a range of BMDL01 values of 0.3 to 8 µg iAs/kg b.w./day for skin lesions and cancers of the skin, urinary bladder and lung. It is the EU reference assessment for inorganic arsenic in food until the 2021 chronic-exposure update at efsa2021-chronic-exposure-iAs.
• Speciation discipline: the opinion is the most thorough European-level statement to date that risk assessment of arsenic in food requires speciation (iAs vs tAs vs organic species), not total-arsenic measurement. Roughly 98 percent of the 100,867 records submitted reported only total As, forcing the Panel to model nine iAs scenarios from literature-derived iAs/tAs ratios. The conclusions section recommends that future official food control should report inorganic arsenic directly, not total arsenic, for food categories where iAs is the relevant species.
• Occurrence base: the 77,275-record cleaned EU occurrence dataset (2003–2008) and the per-category adjusted means (Table 21) are a high-confidence input for ingredient-level arsenic contamination profiles across cereals, rice, fish, seafood, vegetables, coffee, tap and bottled water, and food for special dietary uses when synthesis runs.
• Subgroup exposure: high consumers of rice from ethnic-pattern diets (~1 µg iAs/kg b.w./day) and high consumers of algae-based products (~4 µg iAs/kg b.w./day) sit at the upper end of the BMDL01 range and warrant explicit treatment in any subgroup risk analysis. Infants exclusively fed milk-based formula reconstituted with average European tap water are exposed to iAs at roughly 3 times breast-fed infants but below the BMDL01 range; infants substituting rice-based weaning foods reach 1.6 µg/kg b.w./day on a single rice-food serving pattern.
• Children: dietary iAs exposure in children under 3 years was estimated at 0.50 to 2.66 µg/kg b.w./day, two to three times adult per-kg rates; this is the load-bearing finding for the prioritisation of infant- and child-food product categories in any downstream regulatory or certification work.
• Microbiome: not specifically addressed in this opinion; broader arsenic-microbiome literature enters through other ingest tracks.

Provenance notes

License class public-reference-only. EFSA scientific opinions are publicly available but not licensed for redistribution; the raw PDF is held privately under raw/reports/ and the access_url points to the EFSA Journal landing page for verification. The DOI 10.2903/j.efsa.2009.1351 resolves through doi.org to the same record. A second copy of the same opinion in raw/Manual Fetch Kimi /June 3 Folder/EFSA Journal - 2009 - - Scientific Opinion on Arsenic in Food.pdf was recorded as a near_duplicates entry during the 2026-06-03 merge-enhance; content was spot-checked across pp. 1–10, 19–26, 31–40, 52–75 and 133–148 against the canonical PDF and matched on all numerics, footnotes and table values.

The published EFSA file is the corrected version dated 1 February 2010, which fixed an error in the interpretation of Rahman et al. (2006a) total number of cases (Table 41 and Appendix). The corrected version is the only one currently distributed by EFSA; the original 22 October 2009 version is available on request. The corrections did not change the overall conclusions of the opinion.

Subsequent EFSA work on inorganic arsenic exposure includes the 2014 dietary exposure update (EFSA Journal 2014;12(3):3597), the 2021 chronic-exposure update establishing a new reference-point range based on additional epidemiology (efsa2021-chronic-exposure-iAs), and the 2024 risk-assessment update (efsa2024-risk-assessment-iAs-update).

Verification notes

2026-06-03 — Merge-enhance via /ingest-next-manual-fetch-pdf from raw/Manual Fetch Kimi /June 3 Folder/EFSA Journal - 2009 - - Scientific Opinion on Arsenic in Food.pdf. The June 3 PDF is the same EFSA Journal 1351 opinion already on file at raw/reports/scientific-opinion-on-arsenic-in-food-2009.pdf (DOI match; SHA-256 differs because the two PDFs were downloaded at different times and the EFSA host serves slightly different PDF metadata, but the rendered content is identical on spot-checks of pp. 1–10, 19–26, 31–40, 52–75 and 133–148). Recorded under near_duplicates. The prior page version (updated: 2026-04-25) carried only the abstract-level summary with seven Key numbers, no Methods, no Implications, no Provenance, no Verification. Schema fixes applied: added raw_handle: MFK_efsa-journal-2009-scientific-opinion-on-arsenic-in; added matrices: [dietary-intake] (clears the routing_malformed advisory); updated sampling_year_range from null to [2003, 2008] (the period EFSA used for the cleaned dataset); expanded ingredients from 5 to 10 entries to cover coffee, cereals, vegetables, seaweed and breastmilk (all called out as significant iAs contributors or exposure sources in the opinion); added near_duplicates block; removed legacy ## Summary heading and the legacy ## Wiki pages updated on ingest list (current schema uses un-headed opening prose under H1 and does not maintain page-output bullets on source pages). No prior source-content claims required revision; the opinion’s headline numbers (0.13–0.56 LB to UB µg/kg b.w./day mean adult iAs exposure, 0.37–1.22 P95, 2 to 3-fold child-to-adult ratio, BMDL01 range 0.3–8 µg/kg b.w./day) were already correct and have been verified against the corrected PDF.

Source-text re-verification of every numeric value added to Key numbers: TWI/BMDL ranges from Section 9 (“Conclusions”, p. 142) and Table 43 (p. 141); adult exposure ranges from Section 7.2 (p. 61) and Table 23 (p. 62); iAs scenarios from Table 24 (p. 64), Table 25 (p. 65) and Table 26 (p. 66); child exposure from Table 28 (p. 69); infant exposure from Table 27 (p. 67); specialised-diet increments from Table 31 (p. 74); occurrence means from Table 21 (p. 53); occurrence dataset composition from Section 5.2.1 (p. 34) and Section 5.2.3 (p. 37); iAs/tAs proportion assumptions from Section 5.5 (p. 56–57); food-group contributors from Section 7.3.1 (p. 66) and Figure 10 (p. 67).

2026-06-03 — Audit subagent (general-purpose, fresh context) verdict REVISE. Three ❌/⚠️ findings independently verified against the PDF and applied; four ⚠️ findings independently verified against the PDF and found to be false positives (not applied). Applied:

• Body prose “22 October 2009 adoption” corrected to distinguish adoption (12 October 2009 per Title-page footnote 1) from original publication (22 October 2009 per Title-page header text); the corrected file dated 1 February 2010 replaces the publication, not the adoption.
• Key numbers row “Records with detailed speciation (As species individually quantified) | 174 (0.2 percent)” corrected by relabel + split: the 174 figure refers to organic-arsenic analyses per §5.2.2 p. 36 (“analysis of inorganic and organic arsenic was performed on only 919 (1.2 %) and 174 (0.2 %) of the samples, respectively”); detailed species-level speciation (arsenite/arsenate/arsenobetaine/DMA/methylarsonate individually) was carried out on 158 samples per §5.2.2 p. 37 paragraph 1. Now reported as two separate rows.
• Table 27 Tap water row parenthetical “(formula reconstitution)” dropped — Table 27 p. 67 lists Tap water at 600 g/day as a separate food item without the reconstitution attribution; formula-reconstitution water is folded into the milk-based-formula occurrence figure (1:9 dilution applied inside the formula row).
• §5.5 iAs/tAs row “Up to 60 percent inorganic in hijiki (FSA 2004)” corrected — the FSA 2004 hijiki numbers on p. 33 (110 mg/kg total / 77 mg/kg iAs) imply ≈ 70 percent inorganic, not 60. The “up to 60 percent” figure applies to Spirulina tablets per Almela 2006 cited on the same paragraph of p. 33 and is also generalised to hijiki in §5.5 p. 57. Row now reports both attributions correctly.

Rejected as false positives (verified against PDF and found to be present in the source):

• “tea-infusion iAs varies 29–88 percent of total (Yuan 2007)” — verified on §5.5 p. 57: “A recently published study reported that inorganic arsenic derived from tea infusions can vary from 29 to 88 % of the total arsenic (Yuan et al., 2007).”
• Jorhem 2008 rice cells (brown 0.240, parboiled white 0.210, white 0.100; iAs 64 percent of total) — verified on §5.1 p. 32: “In a Swedish study, the mean concentration of total arsenic in long grain brown rice of 0.240 mg/kg was similar to that of parboiled white rice at 0.210 mg/kg, whereas white rice contained considerably less arsenic (0.100 mg/kg). The concentration of inorganic arsenic averaged 0.110 mg/kg, or 64 % of the total (Jorhem et al., 2008).”
• “Italian INRAN 1994-1996 national survey of 1,940 subjects” — verified on §6.2 p. 59: “the food consumption information available from the 1994-1996 national survey of 1940 Italian subjects carried out by the Italian Istituto Nazionale di Ricerca per gli Alimenti e la Nutrizione (INRAN)…”
• Table 28 child-exposure P50 country-median LB column for ages 3, 6, 10, 14 — audit flagged ⚠️ as “could not be confidently re-verified due to dense table layout”; values in the wiki page were typed from the PDF Table 28 image during ingest and the three matching columns (P50 max, P95 median, P95 max) align cleanly to the source rows, so the P50 median LB column is read from the same horizontal alignment. Spot-check on future audit pass remains advisable but no error is asserted.

Audit Checks 2 (slug vocabulary), 3 (speciation and methods), 4 (Part 12 brand firewall) and 5 (Part 2 wiki/HMTc firewall) were all returned clean by the subagent — no findings to apply or reject in those checks.

Page history

The five most recent substantive edits to this page. The full version history lives in git; when DOI minting comes online (see schema docs), each entry below will also link to a version-pinned DataCite DOI.